MAFE Working Paper 2

Political turmoil, economic crises, and international migration in DR Congo

Evidence from event-history data (1975-2007)

SCHOUMAKER Bruno, Université catholique de Louvain

VAUSE Sophie, Université catholique de Louvain

MANGALU José, Université de Kinshasa

April 2010

The MAFE project is coordinated by INED (C. Beauchemin) and is formed,

additionally by the Université catholique de Louvain (B. Schoumaker), Maastricht

University (V. Mazzucato), the Université Cheikh Anta Diop (P. Sakho), the Université

de Kinshasa (J. Mangalu), the University of Ghana (P. Quartey), the Universitat

Pompeu Fabra (P. Baizan), the Consejo Superior de Investigaciones Científicas (A.

González-Ferrer), the Forum Internazionale ed Europeo di Ricerche sull’Immigrazione

(E. Castagnone), and the University of Sussex (R. Black). The MAFE project received

funding from the European Community’s Seventh Framework Programme under grant

agreement 217206. The MAFE-Senegal survey was conducted with the financial

support of INED, the Agence Nationale de la Recherche (France), the Région Ile de

France and the FSP programme 'International Migrations, territorial reorganizations and

development of the countries of the South'. For more details, see:

http://www.mafeproject.com

1

ABSTRACT

This study documents the impact of political and economic crises in DR Congo on

international migration since the mid 1970s. Using a recent longitudinal survey

conducted in Kinshasa and event history models, we show that political and economic

troubles have influenced migration from DR Congo. Separate models by destination

show notably that political crises have had a strong impact on migration to developed

countries, while both political and economic troubles influenced migration to African

countries. Finally we show that migration is less selective with respect to education in

times of political crises.

1. Introduction

Since its independence in 1960, DR Congo 1 has been hardly hit by econom ic and

political crises. The deteriorating economic context and the wars that h ave affected the

country over the last decades epitom ize the complex situations faced by many African

countries. P olitical instability in DR Congo has generated large flows of refugees to

neighboring countries2, illustrating the impact of crises on international migration. Since

the 1980s, m igration flows from DR Congo to Europe (measured with immigration

statistics and asylum claims) have also increased significantly (Migration Policy

Institute, 2007) and are thought to be reflecting the deteriorating economic and political

situation in the country (Sumata, 2002a).

Although the existing data suggest a link between deteriorating political and

economic conditions and international migration, the impact of crises on international

migration in DR Congo has received very little attention in the scientif ic literature. One

reason may be that this impact seems obvious. Yet, research in other contexts (mainly in

2

Latin America) has shown that the influence of political and economic crises on

migration may be complex, and that crises do not necessarily increase international

migration. Previous research also suggest that economic or political crises may

influence migration to different destinations in different ways (Massey and Capoferro,

2006; Jokisch and Pribilsky, 2002). Likewise, crises can influence the composition of

the flows of migrants or, said dif ferently, can have different effects on different

categories of people (Massey and Capoferro, 2006; Jones, 1989).

The lack of studies on this topic also refl ects the lack of ap propriate data. Data on

migration flows from DR Congo to African countries are very limited (not to say

inexistent). Data on ‘stocks’ of refugees (UNHCR, 2010a), and on the variations in

stocks of refugees (Moore and Shellman, 2004) provide useful inform ation to analyze

trends in m igrations. However, they cover refugees and refugee-like people (UNHCR ,

2010b), but do not include other categories of m igrants. Data on refugees may also be

affected by serious measurement error 3 , and often do not include individual

characteristics (age, sex…) of refugees.

In Western countries, existing data on Congol ese migration mainly include data on

flows of Congolese im migrants and asylum se ekers. W hile these data are also very

useful, they have several drawbacks for an alyzing the im pact of crises on m igration

from DR Congo (and from developing countries more generally). First, the analyses are

conditioned by the availability of data in the destination countries, and such data are not

readily available in many countries.4 Lack of data on undocumented migrants is another

limitation. In addition, the dates recorded in the statistics on imm igration and asylum

demands may not correspond to the dates of de parture of the countries of origin, and

may not be appropriate to analyze the re lationship betwe en cris es a nd m igration5 .

3

Finally, aggregate data on migration flows do not allow analyzing differential migratory

responses according to individual characteristics.

The aim of this study is to docum ent the impact of political and economic crises in

DR Congo on in ternational m igration, us ing a recent longitudin al su rvey on

international migration conducted in Kinshasa (MAFE-Co ngo 2007)*. These individual

data collected in the or igin area a llow more detailed an alyses than the aggreg ate data

presented above. Using event history m odels, they are com bined with m acro-level

variables to m easure the im pact of poli tical and econom ic cris es on m igration (see

details in data section).

This paper is organized around three objectives:

(1) The first objective is to measure the impact of economic and political crises on

the risks of international m igration. Even though the impacts of economic and political

changes are not easily disentangled in migration analysis (Morrison, 1993), we take into

account both political and econom ic m acro-level variables to distinguish their

influences on migration.

(2) The second objective is to test if political and economic crises have had similar

effects on migrations to African cou ntries and to Western countries (Europe and North

America). Existing data make it dif ficult to e stimate if some destinations are pref erred

during crisis. W hile aggregate data on i mmigration and asylum claim s in European

countries suggest that European countries are particularly attractive in tim es of crises,

such conclusions do not take in to account the fact that m igration to other destinations

may also increase.

4

(3) The third objective of this study is to evaluate if economic and political changes

influence different people in different ways . More specifically, we test the hypothesis

that m igration is less selective with resp ect to education in tim es of econom ic and

political tro ubles or, said dif ferently, that the impact of crises is s tronger on the less

educated than on the more educated people.

2. The Political and Economic Context in DR Congo

The Democratic Republic of Congo is one of the largest countries in Sub-Saharan

Africa. It is the second larg est Sub-Saharan African country in term of area (after

Sudan), and with a population es timated at 59 millions in 2005 (United Nations, 2009),

DR Congo is also the third m ost populated country in S ub-Saharan African (after

Nigeria and Ethiopia), and the largest coun try in francophone Af rica. DR Congo is

currently one of the poorest countries in the W orld. According to the Hum an

Development Index (UNDP, 2009), it ranked 177 out of 179 countries in 2008.

Since the country gained independence in 1960, Congo has experienced a series of

economic downturns and of episodes of po litical instability. Overall, seven broad

periods can be distinguished in the c ountry’s political and econom ic history ( Erreur !

Source du renvoi introuvable.). The period from 1960 to 1965 (First republic) started

with independence from Belgium and ende d with the seizing of power by Joseph

Mobutu (Stengers, 1989; Ndaywel, 1998; Ba mba, 2003). That period was a tim e of

political ins tability, in w hich several partie s struggled for power, m ost of which were

constituted on regional or trib al grounds (Putzel et al., 2008 ). From the economic point

of view, the period was characterized by a stagnating econom y, hi gh inflation, and a

deterioration of external balance (Peem ans, 1997; RDC and UNDP, 2 000). However,

the repercussions of the econom ic troubles on the population are t hought to have been

5

limited at that tim e, notably b ecause infras tructure and social system s were still

operating (Ndaywel, 1998).

-- Table 1 about here --

The Mobutu regim e started at the end of 1965, and opened a new era in Congo’s

political and economic history (the S econd Republic), which lasted from 1965 to 1997.

At first the political s ituation was fairly stab le, and the country’s econom y al so

improved at the beginning of the Second Republic. Between 1965 and 1974, the average

GDP growth rate was around 7% per year (Nzisabira, 1997; RD C and PNUD, 2000).

The high prices on the world c opper market, the increase of fo reign direct investments,

and the growing internal m arket all contri buted to the positive econom ic performances

(Peemans, 1997; Nzisabira, 1997). However, governm ent spending and debts also

increased during that period, notably to fund projects of dubious profitability (Peemans,

1997; Nzisabira, 1997; RDC and PNUD, 2000).

Between 1975 and 1982, the econom ic situation seriously deteriorated. The 1973

oil crisis, along with the collapse of the price of copper (1974) and of ot her

commodities, and bad econom ic policy (no tably the process of zaïrianisation of the

economy which started in 1973), ruined the be nefits of the preced ing period (Peemans,

1997; Nzisabira, 1997). Foreign investm ents decreased, public expe nditures increased,

and the country accumulated a large foreign debt, in unfavourable conditions (RDC and

PNUD, 2000). From 1975 to 1978, the GDP dropped by 3.5 percent annually. T hat

period was also characterized by political turmoil. Opponents to the regime started wars

to try gain control of the country, notab ly in 1977 and in 1978 (Shaba wars). The

6

support of Western countries was decisive to control the rebellions (Peem ans, 1997;

Ndaywel, 1998).

The period from 1983 t o 1989 started with economic reform s, and with a slight

increase of the GDP, along with the start of a structural adjustment program with the

World Bank and the IMF (Nzisabira, 1997; RDC et PNUD, 2000). In 1986, the country

broke off the negotiations with Bretton W oods institutions, and the GDP growth rate

plunged (Nzisabira, 1997; Mutam ba Lukusa, 1999). Inflation increa sed (from 65% in

1986 to 75% in 1987), and the devaluation of the currency accelerated (Peemans, 1997).

At the end of that period, the GDP growth rate was negative (-1.3 % in 1989). This also

coincided with the end of the cold war and with serious changes in the political situation

in DR Congo.

The 1990s is one of th e darkest p eriods in Congo’s recent political and economic

history. The com bined pressure of the in ternal opposition and of the internation al

community forced the President M obutu to announce (in April 1990) the

democratization of the regim e. However, th e democratization proc ess not only lagged,

but was als o accompanied by political in stability. Riots eru pted in 199 1, shortly after

the Conférence Nationale Souveraine was set up to dec ide the f uture of the cou ntry

(Hesselbein, 2007). The political situation was very tense and unstable. In 1994, the

genocide in Rwanda leads to m assive fl ows of people from Rwanda to DR Congo,

mainly refugees but also “members of the former Rwandese Government along with the

army and m ilitias wh o had orga nized and comm itted the genoc ide (Hesselb ein,

2007: 38). This contrib uted to political instability in DR Congo. In 1996, with the

support of Rwanda, Uganda and Angola, Laur ent-Désiré Kabila declared war on the

Mobutu regime (McCalpin, 2002). By May 1997, Mobutu had fled to Morocco and the

7

rebellion led by Laurent-Dési ré Kabila (AFDL) had se ized power. In 1998, “Rwanda

and Uganda orchestrated another rebellion in eastern Congo – this time with the goal of

deposing the man they had imposed a year earlier” (Dunn, 2002: 63). This was the start

of the second Congo War, which was to last until 2003.

Laurent-Désiré Kabila rem ained in power until he was assassinated in January

2001. He was replaced by his son Joseph, who quickly m ade overtures to the

international comm unity, with the willingn ess to m ove towards peace (Putzel e t al.,

2008). From 2002, violence was significantly reduced after agreements were signed by

the belligerents (Hesselbein, 2007), and the second Congo War officially ended in 2003.

In 2006, elections were organized and Joseph Ka bila was elected President. Despite the

end of the war, eastern Congo has continued to be regularly prone to violence.

The 1990s were also characterized by a rapid deterioration of the econom ic

situation. The estimated GDP growth rate, which was already negative at the beginning

of the period, decreased from -6.6% in 1990 to -14% in 1999. Congo’s econom y was

struck by hyperinflation, and the country’s public debt also soared. Official

Development Aid also decreased drastically in the 1990s (Hesselbein, 2007; Mutam ba

Lukusa, 1999). During that period, the purch asing power of the Congolese population

declined considerably.

Since the year 2000, and especially sin ce 2003, the econom ic context and living

conditions of the population have slightly improved. In 2002, the country experienced

positive GDP growth rates for the first tim e since 1995. The im provement in the

economic situation is explained by the post -war reunification of the country, by the

resumption of international development aid, the control of public finances, and a

8

massive injection of foreign currency by the IMF (IMF, 2002). However, the living

conditions of the Congolese population remain extremely difficult.

As shown by this overview, Congo’s hist ory has been characterized by large

economic and political changes. It is also clear that, in DR Congo as elsewhere,

economic and political troubles are clos ely intertwined (Akokpari 1999; Sm ith 1992),

which makes it difficult to disentangle their effects on migration. However, political and

economic conditions are not perfectly correlat ed, and results of event history m odels

shed light on their respective effects on international migration.

3. Congolese Migration over the Last Decades

As noted by several researchers, Congol ese m igration has been little studied

(Ngoie Tshibam be and Vwakyanakazi, 2008) . In W estern countries, Congolese

migration has mainly been studied using imm igration statistics and statistics on asylum

seekers. In African countries, data sources are even scarcer, and quantitative research on

Congolese m igration to other African countries are not readily available, apart from

some research on m igration to South Africa (S teinberg 2005). Despite the lack of data

and studies, it is possible to draw a general picture of Congolese migration.

Congolese migrations within Africa are to a large extent dire cted to neighbouring

countries. F or instance, Angola an d Congo Brazzaville are m ajor destin ations for

migrants originating from W estern Congo, where Kinshasa is located (Schoum aker,

Mangalu and Vause, 2009), while Zambia is a common destination for migrants coming

from Katanga in South-Eastern Congo (N goie Tshibam be and Vwakyanakazi, 2008).

Congolese m igrations to neighbouring countries also include m ovements of refugees,

which have been quite large since Congo’ s independence. In 2008, according to the

UNHCR (2010a), m ore than 200 000 Congolese refugees were living in neighbouring

9

countries, most of the m in Tanzania, Za mbia, Rwanda and Uganda. Since the 1990s,

South Africa has also b ecome a m ajor des tination for Congolese m igrants (Steinberg,

2005; Sumata, 2002). It started increasing in 1989 and intensified in the early 1990s as a

result of econom ic uncerta inty (Steinberg, 2005) as well as political instability and

lootings (Kadima 1999, Ngoie Tshibam be and Vwakyanakazi, 2008; Sum ata, 2002b).

Between 1994 and 2004, around 25 000 Congolese ap plied for refugee status in South

Africa (Crush, 2008). More than 20 000 Congolese were registered as refugees in South

Africa in the early 2000 s, and many more were living without refugee status (Steinberg

2005).

Congolese migration to Western countries has also increa sed significantly over the

last 30 years, accord ing to m igration sta tistics in selected European Countries.

Congolese m igration to Europe started in the early 1960s, after Congo gained its

independence from Bel gium. At that tim e, m igration to Europe was prim arily a

migration of elites moving to Belgium for training (Kagne and Martiniello, 2001; Ngoie

Tshibambe and Vwakyanakazi, 2008). In the 1980s, econom ic m igration gained

momentum and, since the 1990s, asylum -seekers have represented a very large

proportion of Congolese m igrants in Europe. Over the past 30 years, the profiles of

Congolese migrants and their destinations ha ve also progressively diversified (Sum ata

2002a; Ngoie Tshibam be and Vwakyanakazi, 2008). W hile Congolese m igrants were

mainly constitu ted of elites before the 1980s, less educated and poorer peop le are

thought to have increasingly participated in international migration to Western countries

(Sumata, 2002a). Destinations have also changed: France becam e an increasin gly

popular destination among Congolese and, more recen tly, the United Kingdom and

Germany have attracted a siz eable share of the Congolese m igrants in Europe (Sum ata,

10

Trefon and Cogels, 2004). Other destinations such as Canada and the United States have

also becom e increasingly popular am ong Congolese m igrants, but the num ber of

Congolese m igrants living in these countries rem ains much lower than in Europe.

Currently, the largest comm unities of Congoles e migrants in W estern countries liv e in

France (approximately 90 000 Congolese m igrants in the early 2000s, see Bazenguissa-

Ganga, 2005) and in Belgium (a pproximately 50,000 Congolese m igrants in 2007,

Schoonvaere, 2009).

Although data and research po int to increasing num bers of Congolese m igrants in

African, European and North Am erican countries, especially since the 1990s, there is a

lack of res earch on the tim ing of migrations and on the ch aracteristics of m igrants. If

one wants to link increasing m igration with economic and political troubles, tim e series

on migration are needed. In Af rican countries, statistics on migration flows are lacking,

and to our knowledge, no study has been done on Congolese migration trends to African

countries. In Europe, imm igration statistics and st atistics on asylum seekers provide a

more detailed picture of vari ations in Congolese m igration over the last decades. T he

available data suggest that m igration from DR Congo to Europe was especially intense

during the periods of crisis in DR Congo si nce the early 1990s. A ccording to Sum ata

(2002a: 625), the “unp recedented level of po litical and so cioeconomic crises was a

fertile ground for continuing [Congolese] immigration [to E urope]”. However, data on

annual migration flows to Europe m ay be diffi cult to interpret for several reasons: lack

of data on undocum ented migrants, time lag between departure from Africa and arrival

in Europe, lack of infor mation on place of depa rture of migrants. As stated before, lack

of data on the characteristics of m igrants also hamper detailed analyses of Congol ese

11

migration, and of the differential im pacts of crises according to individual

characteristics.

4. Crises and Migrations: A Brief Review of the Literature

To date, the impact of macro economic and political conditions on migration has,

surprisingly, rece ived r elatively little attention in developi ng countries. From a

theoretical point of vie w, m any authors r ecognize that m acro-economic and political

conditions are m ajor driving factors in m igration. For instance, in the African context,

Adepoju (1994) m entions four types of m acro conditions that can explain m igration

dynamics: the deterioration of economic conditions, changes in demographic population

(which increase underem ployment), politic al instab ility and cultural p ractices

(traditional practices tend to push young people to leave the country). According to him,

economic conditions and political instability are the two m ost important factors behind

the flows of m igrants and refugees. Yet, em pirical analyses on this topic, especially in

sub-Saharan Africa, are quite scarce.

Migratory responses to wa rs and political turm oil (including m ovements of

refugees and asylum -seekers) have been docum ented in places like ex-Yugoslav ia

(Conti and Mam olo 2007), the Gulf Count ries (Addleton, 1991; Russell, 1992) and

Vietnam (Merli, 1997). Moore and Shellm an (2004: 723) also used a global sam ple of

countries over 40 years to m easure the im pact of violence (wars, dissident violence…)

on forced m igration, and concluded that “violent behavior has a substantially larger

impact on forced m igration than variables such as the type of political institution or the

average size of the econom y”. All in a ll, po litical turm oil gene rally has a positive

impact on migration, although – as we shall see later – this general conclusion should be

qualified.

12

Most of the em pirical litera ture on the im pact of ec onomic crises in departure

countries on migration has been done on Latin Am erican countries6. A recen t paper by

Massey and Capoferro (2006) st udied the impact of the de teriorating economic context

on international m igration in Peru in the 1980s and the 1990s. Using longitudinal data

of the Latin Am erican Migration Project (LAMP) and event history m odels, they

showed that the start of the structur al adjustm ent programm e, along with the

deterioration of e mployment opportunities a nd wages, coincided with an increase in

international m igrations. They also showed a diversification of the destin ations of

migrations (with an increase of m igration to European countries), as well as a decrease

in the selectivity of m igration. Their work indicates that, before the econom ic crisis,

migration was m ore selective w ith respect to education. T heir explanation is that, in a

‘reasonably functioning labor market’, people are likely to move to maximize earnings,

and will b e positively selected with resp ect to edu cation, as exp ected from the

neoclassical theoretical pe rspective (Massey and Capofe rro, 2006). On the other hand,

in periods of econom ic downturns, peopl e tend to flee deteriorating economic

conditions rather than seek to m aximize earnings abroad, and m igration becomes l ess

selective. T he authors suggest this type of be havior is more in line with the new

economics of labor migration, according to which people move to overcome missing or

failed markets.

Work was also done in Ecuador on this topic (Jokisch and Pribilsky, 2002),

although the impacts of economic and political troubles on migration were not tested i n

statistical models. The authors consider that new Ecuadorian m igrations are a response

to economic and political crises. Before th e 1990s, Ecuadorian inte rnational migrations

were m ainly directed to the United Stat es. From the 1990s, Ecuador experienced a

13

political and economic crisis, which coincided with stronger immigration policies in the

US. The authors observe increasing m igration flows in tim es of cr isis, a change in

destinations (m ore m igrations to Eu ropean coun tries lik e Spain, France, Italy ), and a

diversification of m igrants profiles in term s of gender (fe minization) and socio-

economic status. Contrary to results found by Massey and Ca poferro in Peru, Jokisch

and Pribilsky (2002: 91) suggest that migration to the United States are m ainly done by

“poor campesiños from the countryside”, while “ migration to Spain is a phenom enon

capturing the imagination of Ecuadorians of all classes”. In s hort, migration is also less

selective in times of crises, but in a different way from what is found in Peru.

Little research has looked sim ultaneously at the effects of ec onomic and political

troubles on migration. One of the early works on this topic was done by Stanley (1987),

who studied inte rnational m igration f rom Sa lvador to th e United Sta tes in th e ea rly

1980s, and tried to distinguish the im pact of political and econom ic factors. Using

aggregate tim e series data on m igration fl ows (US Imm igration and Naturalization

Service app rehension s tatistics) an d indicators of econom ic and political troub les

(political violence), Stanley (1987: 147) concluded that "fear of political violen ce is an

important and probably the dom inant motivation of Salvadorans who have m igrated to

the U.S. since the beginning of 1979”. In c ontrast, economic factors were considered

less im portant, although the author underlines that econom ic conditions interact with

political turmoil in a num ber of ways, e .g. violence can “disrupt econom ic activities,

thereby eliminating jobs and reducing pay levels” (Stanley, 1987: 133). Also working

on m igration from Salvador to the Unite d S tates, Jones (1989) reached different

conclusions. According to him, bad economic conditions did influence migration to the

United States. A spatial analysis of regions of origin of migrants to the United States led

14

him to conclude that “politic al violence has m ost affected the relatively poor Northern

provinces, but a lack of money and knowledge m akes flight to the Un ited States out of

the question” (Jones, 1989: 194)7.

All in all, the existing resea rch remains scarce, is overall not very recent and has

led to m ixed conclusions. The general idea is that deteriorating econom ic and political

conditions tend to increase migration, although this should be qualified, notably because

the im pact can vary accordi ng to the destination of m igration and according to the

characteristics of individuals (e.g. educat ion and gender). Resear ch on the relative

importance of political and economic crises has also provided mixed results. Finally, the

impact of econom ic and political crises on Af rican m igration is alm ost totally absent

from the literature.

In this paper, we treat the three following questions:

Question #1: Do economic and political troubles increase migration? Our hypothesis

is that, as economic conditions deteriorate, people will move to neighboring countries or

more distant countries to secure a job or better wages. As stated by the New Econom ics

of Labour Migration, m igration can be view ed as a risk-dim inishing strategy in a

context of m arket failures (Massey and Ca poferro, 2006). As lo cal possibilities of

diversification are scarce in DR Congo, international m igration is a m ajor way of

diversifying sources of incom e in a cont ext of econom ic deterioration, hyperinflation

and lack of credit and insurance m arkets (Sumata, 2002a) . We also expect m igration to

increase in times of political troub les. People may flee to neighboring countries during

wars8. Another reason why peopl e moved during periods of political troubles m ay be

due to the regime change, especially among people who used to be close to the Mobutu

regime (Sumata, 2002a).

15

Question #2: Do the impacts of economic and political troubles on migration vary by

destination? As discussed in the literature revi ew, economic and political crises m ay

have different im pacts depending on the dest ination. W e expect that the im pact of

deteriorating econom ic conditions will be larger on m igration to Africa than on

migration to Europe. Our hypothesis is that , as the econom y deteriorates, financial

resources to m ove to distant countries ar e less readily available, and people m ay be

tempted to move to close countries with be tter economic prospects (e.g. South Africa,

Angola). Although the im pact of political crises on m igration m ay also vary by

destination, the differences are not obvious a prio ri. Fleeing to clos e countries m ay be

easier, and as a result it m ay be expected that m igration from Kinshasa to neighbouring

countries would increase m ore than migration to Western countries. On the other hand,

Western countries may be viewed as a safer alternative, and migration to these countries

may increase to a greater extent in periods of political troubles.

Question #3: Is migration more or less selective according to education in times of

crises? According to Massey and Capoferro (2006) , migration is less s elective in times

of econom ic cris es. T heir argum ent can be s ummarized as follows: in periods of

economic downturns (and probably in periods of political instab ility), people tend to

flee deteriorating econom ic conditions rather than to m aximize earnings abroad. As a

result, the selectivity should be lower in peri ods of crisis than in periods of econom ic

stability. A similar argument was also put forward by Sumata (2002a) in the case of DR

Congo. According to him, in the 1990s, “both rich and poor had no choice but to seek

political asylum ” (Sum ata, 2002a: 625). In contrast, Jones’ work on m igration from

Salvador to the United States (Jones, 1989), has pointed out th at the poorest people

were not able to m igrate to the United Sta tes to f lee violence in perio ds of politic al

16

crises. In other words, m igration could be m ore selective with resp ect to educ ation in

times of crises.

5. Data

This study relies on retrosp ective data collected in Kinshasa (DR Congo) in August-

September 2007 as part of a research project called “Crisis and in ternational migration

in DR Congo”. This project is a collaborative research project between the University of

Kinshasa (DR Congo) and the University of Louvain (Belgium ), and is part of an

international research program on ‘Migration between Africa and Europe’ (MAFE9).

The survey was conducted am ong a repres entative sam ple of 943 households in

Kinshasa, the capital city of DR Congo 10. Selected retrospective data were collected in

the household survey, and full life histories were also collected from 992 adults in these

households (males and females, return migrants and non-migrants) aged between 20 and

60. Only data from the household questionnaire are used in this paper.

In all the households, questions were aske d to identify all the people who had lived

in the household at som e point in tim e a nd who had gone abroad for at least three

months, whether they were still living abroad or had returned to DR Congo. In addition,

brothers and sisters of the household head and of his/her spouse who had lived out of

DR Congo were also identifie d through this questionnaire. Data on the m igrations of

these individuals (year of departure, destination country, year of first return if the person

returned, etc.) were collect ed. In addition, socio-dem ographic characteristics (age,

gender, education, marital status and date of marriage, place of birth) were recorded for

all the current m embers of the households as well as for the m igrants. The availability

of data for both m igrants and non-m igrants, as well as data on the tim ing of m igration

enables the use of event history models. In this research we use data on current

17

members of household and past members of households (data on brothers and sisters are

not used, unless they were part of the household at some point in time).

Macro-level data are used in this r esearch to m easure changes in econom ic and

political conditions. Two types of data are used. Political con ditions are captured us ing

an index of political troubles. The index is com puted using the “Internal W ars and

Failures of Governance, 1955-2007” data set p repared by the Political In stability Task

Force (PITF) at the Sch ool of Public Policy (George Maso n University) 11. The dataset

includes four types of political instability events for all the years since 1955. These

events include “ethnic wars, revolutionary wars, genocides and politicides, and adverse

regime changes. Each annual record for each event includes three m easures of

magnitude and a com posite m agnitude score” (PITF 2009). 12 Our ind ex of political

troubles uses all four types of events, and is computed as a weighted average of the four

composite magnitude scores, using principal components analysis. The first component,

which explains 68% of the variance, is used as the index of political troubles. The value

of this indicator is shown on Figure 1. Economic conditions are measured with the GDP

growth rate. The GDP data were obtaine d from the World Developm ent Indicators

online data base (W orld Bank 2009). In the ev ent history models, the average of these

(standardized) indicators for the two preceding years is used. This is b ased on the id ea

that migrants respond with a time lag to deteriorating economic and political conditions,

both because some time may be needed to realize that the situation is deteriorating, and

some time may also be required to decide and organize the migration.

These indicators suffer from some limitations. For exam ple, the GDP growth rate

is affected by m easurement errors . Also, the GDP does not necess arily m easure

precisely the deterioration of the living conditions of th e population, especially in a

18

country with a large inf ormal sector. However, it is one of the few indicators for which

time series are availab le, and which reas onably describes econom ic conditions in the

country. The index of political troubles also has some limitations. One of them is that all

the conflicts are taken into acc ount in the indicator even if they do not directly affect

Kinshasa. H owever, we believe this indicato r provides a more deta iled pictu re of the

intensity of political troubles in the country than simple dichotomous indicators.

6. Methods

Event history models are used to reconstr uct migration trends and to evaluate the

impact of political and economic crises on migration since 1975 among people aged 15.

Event history m odels are particularly we ll-suited to study the im pact of sudden

economic and political changes on m igration, as tim e-varying variables at m icro and

macro levels can be included in the m odels (Henry, Schoum aker and Beauchem in,

2004).

Piecewise exponential models are used (Allison, 1995; Blossfeld et al., 2007).

These models rely on the organization of the da ta file as a person-period file. Each line

in the d ata f ile rep resents a pe riod of tim e during which the explanatory variables

(including age and year) are constant. Th e dependant variable (d ummy variable)

indicates if the even t (interna tional m igration) takes p lace during th e tim e inter val

corresponding to the line in the data file. Th e rate of m igration is supposed constant

within each time interval. Age (in single yea rs) and years are included in the models as

time varying variables, so that in p ractice migration rates are allowed to vary each year

and at each age. The model is estim ated with Poisson regress ion (Allison 199 5;

Blossfeld et al. 2007), and an offset is include d in the models to control for the varying

lengths of the periods (exposure)13 .

19

The population at risk of expe riencing a migration includes all the people currently

living in the households, and people who lived in the household in the past and who

have migrated to another count ry for at least 3 months (whe ther they were still living

abroad at the tim e of the surv ey or had returned to DR Congo). The data set includes

non-migrants, return m igrants, as w ell as m igrants still living abroad. Individuals are

included in the data set from age 15 and from year 1975. For people older than 15 when

entering the data set (1975), this corresponds to a situation of late entry (left truncation)

(Allison, 1995; Guo, 19 93). The an alysis period starts in 1 975 for two reason s. First,

because of the retrosp ective na ture of the m igration data, sam pling errors increase as

one goes back in tim e and as the sample size gets smaller. Moreover, as explained in

section 2, the economic and political situation was fairly stable until the mid 1970s. The

analyses are also restricted to migration after age 15, as we are interested in autonomous

migration14.

The reconstruction of m igration trends is done using the piecewise exponential

model only including age and year as indepe ndent variables. A linear function of age

and logarithm of age is included in the m odel to control for age, and calendar time

(years) is taken into account in the models in two separate ways. First, it is included as a

series of dumm y variables (non param etric approach), in order to m easure annual

variations in m igration risks. Seco ndly, linear splin es a re f itted to s mooth m igration

rates and to identify breaking points in migration trends. The number and location of the

knots (breaking points) are estim ated using a stepwise forward m ethod of selection

(Marsh and Cormier, 2001).15 Results are presented in Section 7.1.2.

20

Piecewise exponential models are also used to test the impact of political and

economic troubles on m igration rates. Three series of models are estim ated, that

correspond to the 3 research questions.

The first s eries of models are fitted to m easure the impact of political and

economic troubles on migration rates (Question #1). The tw o indicators of political and

economic troubles are f irst included separately in the m odels, and are then included

jointly. Individual variables (age, marital status, education) are controlled in the models,

as well as a variable measuring the migration trend (year).

The second series of models is sim ilar to the first series, but distinguishes

migration by broad destination (Europe a nd North Am erica vs. Africa). These m odels

are fitted to test if the impact of economic and political troubles on m igration varies by

destination (Question #2). Sepa rate analyses rely on recodi ng the dependant variable to

take into account the m igration of interest . For exam ple, if one is interested in

migrations to Europe and North Am erica, th e dependant variable is equal to 1 if a

person migrates to Europe or North America, and zero otherwise. Migration to the other

region is considered as censoring.

Finally, the third s eries of models explores the inte ractions between economic and

political troubles and education. More spec ifically, the objective is to identify if

migration is less or more selective with respect to education in times of crisis (Question

#3). Said differently, these m odels aim at test ing if crises im pact different people in

different ways. Models with interactions are e stimated se parately f or m igration to

Western countries, African countries and all countries.

21

7. Results

We f irst pr esent d escriptive resu lts, and next turn to the event history m odels

corresponding to the three research questions.

7.1 Influence of Political and Economic Crises on Migration: Descriptive Analyses

7.1.1 Political and economic crises: 1975-2007

The two indices presented in section 5 are used to m easure changes in econom ic and

political co nditions sin ce 1975 in DR Cong o (Figure 1 ). They illu strate the v ery

changing po litical and econom ic conditions in DR Congo describ ed in section 2. As

shown by the GDP growth rate, econom ic conditions started deterior ating seriously at

the end of the 1980s and rem ained very poor until the early 2000s. The situation was at

its worst in the early 1990s, when the GDP wa s decreasing at a rate close to 10% per

year. Since 2002, the country’s economic situation has im proved significantly but the

situation remains fragile.

-- Figure 1 about here --

The index of political troubles summ arizes the political history of DR Congo since

the m id-1970s. The first hum p corresponds to the 80-day war and the Shaba war in

1977 and 1978. The 1980s were fairly stable from a political point of view. The

situation started deteriorating in the early 1990s with the regime crisis and riots in 1991

and 1992. The start of the war in 1996 and th e regime change in 1997 (replacem ent of

President Mobutu by President L. Kabila) correspond to peri ods of increasing political

troubles. Between 1996 and 2002 (periods of wars) the value of the index rem ains very

high. The war officially ended in 2002, and the political situation improved significantly

from 2003.

7.1.2 Reconstruction of migration trends: 1975-2007

22

Event history models were used to estimate rates of first international migration by age

and year. We use life table m ethodology to tran slate these rates into ‘lif etime risks’ of

doing at least one international migration between age 15 and 50 16. The lifetime risk of

international m igration can be interpreted as the probability of doing at least one

international m igration between 15 and 50. Figure 2 shows annual lifetim e risks of

international migration and linear sp lines fitted to these ann ual risks. The overall tre nd

is clearly an upward trend: the risks were a little over 10% until the mid-1980s, they

went as high as 50% in the early 2000, and decreased to about 30% in 2007. Four knots,

which correspond to breaking points in the m igration trends, were located with the

spline regressions: 1983, 1991, 1996, and 2001. Rate s started to increa se significantly

around 1983; in the early 1990s, the trend stabili zed. Of particular significance is the

large increase in m igration risks after 1996 (starting in 1997, the end of the Mobutu

Regime), a nd the significant decreas e after 2001, when econom ic and political

conditions improved.

-- Figure 2 about here --

7.1.3 Comparisons of political and economic conditions and migration trends

The comparison of the indices of economic and political troubles 17 and migration rates

(Figure 3a and Figure 3b) indicates overall a good consistency between migration trends

and trends in political and econom ic conditions. The consistency is particularly striking

for the index of political troubl es: the large increase in inte rnational migration rates at

the end of the 1990s corresponds with the intense political crisis. T he correlation

between m igration rates and changing econom ic condition s is less striking; however,

this figure shows that since the mid-1990s, periods of deteriorating economic conditions

23

(grey curve going up) correspond to increasing m igration rates, while periods of

economic improvement are accompanied by decreasing migration rates.

-- Figures 3a and 3b about here --

7.2 Influence of Political and Economic Crises on Migration: Event history models

Although g raphical an alyses prov ide s trong hints of a positiv e relationsh ip

between political and econom ic troubles and migration, they do not allow testing the

statistical significance of these relationshi ps, nor do they allow distinguishing the

effects of ec onomic and political conditions, or testing th e differential impact of crises

on different categories of people. In the next section, ev ent history m odels are used to

measure the relationship s between political and econom ic conditions and international

migration from Kinshasa.

7.2.1 Question #1: Do economic and political troubles increase migration?

Models 1 to 3 measure the effects of ec onomic and political troubles on m igration. All

the m odels include age, education, gender and m arital status as individual control

variables (age not shown). Year is also included to capture th e o verall trend in

migration, regardless of the variations in pol itical and economic conditions. In the f irst

model, the index of politica l troub les is included as the only m acro variable. GDP

growth rate is included alone in the sec ond model. The third m odel includes both

variables.

First, indiv idual-level v ariables sho w expected results. M ales a re s ignificantly

more likely to migrate than females. People who have never been married are also twice

as likely to migrate for the first time at each age as married people. Finally, education is

strongly correlated to migration. People with secondary or higher education are more

24

than twice as likely to m igrate as their les s educated coun terparts. Results also sho w a

positive migration trend (variable ‘year’), but th e trend is n ot significant in two of the

three models.

-- Table 2 about here --

When included separately in the models , econom ic and political troubles both

significantly increase the risk of international m igration, controlling for individual

factors (Table 2). In the first model, an inc rease of one standard deviation in the index

of political troubles is associated with an increase of m igration rates of 27%. The

second model shows that that an increase of one standard deviation of the GDP growth

rate is assoc iated with a decrease in migration rate of 11%. W hen both indicators are

included in the sam e model (m odel 3), the impact of econom ic conditions is almost

completely offset and is not significant. The impact of politic troubles, on the other

hand, is very strong and largely s ignificant. Economic and political conditions are not

independent from each other; bu t this resu lt suggests that m igration is m uch more

responsive to political troubles than to economic troubles (in DR Congo, for the time

period concerned).

7.2.2 Question #2: Do the impacts of economic and political troubles on migration

vary by destination?

Separate analyses by destination show seve ral important results (Table 3). Firs t,

as expected , education is a m ajor dete rminant of m igration to Europe and North

America, but much less so for African migration. Rates of first migration to Europe and

North America are more than ten times higher among people with higher education than

among less educated people. Migrations to Africa are, on the ot her hand, m ore likely

25

among people with secondary education. Interestingly, gender is a strong determinant of

migration, but with opposite effects for m igration to African and to W estern countries.

Migration rates to Europe and North Am erica are m ore than 50% higher among

females, while m igrations to Afr ica are m ore frequent a mong males. For both

destinations, single persons are, as expected, significantly more likely to migrate.

With regards to econom ic and political conditions, m odels show contrasting

results by destination. Migrations to Europe and North America seem to depend only on

political con ditions, while econom ic conditions do not influence m igration to Europe

when contr olling f or p olitical trou bles. An in crease of one stand ard devia tion o f the

political troubles index is asso ciated with a 41% increase of m igration rate to W estern

countries. Interestingly, the overall trend measured by the variable ‘year’ is negative: in

other words, after controlling for political troubles in DR Congo, t here is a slight

decrease of m igration rates to Europe and North America. It suggests that the

improvement of political cond itions in DR Congo m ight lead to a stabilization of

migration rates to Europe and North America.

-- Table 3 about here --

Migrations to Africa, on the other ha nd, depend on both econom ic and political

conditions. Both indices have sim ilar (and expected) effe cts on m igration (the GDP

growth ra te is m ore sig nificant): an incr ease of one standard devi ation in political or

economic troubles corresponds to an increase in the rate o f migration of a little more

than 10%. The m odel also shows that ther e is a res idual positiv e tre nd in m igration

rates, after controlling for political and economic troubles. Said differently, migration to

Africa has been increasing, regardless of political and economic troubles.

26

Overall, crises influence m igration to Western countries and to Africa, but in

different ways. Deteriorating economic conditions tend to influence migration to Africa,

but have no clear impact on m igration to Western countries. This suggests that

migration to Western countries – which is m ore costly - is not the preferred strategy in

times of economic crisis. If jobs or better wages are available in neighboring countries,

these destinations seem to be preferred. On the other hands, politic al troubles strongly

influence migration to Western countries. As discussed before, this m ay be due to the

fact that W estern countries are con sidered as s afer destinations in tim es of political

crises. As discussed below, greater legal pos sibilities of migration to Western countries

in times of political crises may also increase the incentive of moving to these countries.

7.2.3 Question #3: Is migration less selective according to education in times of crises?

Previous models have shown that pol itical and econom ic crises influence

migration. Econom ic a nd political troubles bot h influence m igration to Africa, while

migration to Europe and North America is influenced only by political conditions in the

departure country. These models have also shown that migration propensities depend on

individual characteristics. As explained earlier, crises may influence different categories

of people in different ways. The third hypothesis we test in th is section is that migration

became less selective with respect to education in times of crises.

Table 4 shows the results of three models that include the intera ctions between

education and the two m acro-level variab les ( for all des tinations, Africa, Europe &

North Am erica). Th e in teractions b etween education and political conditions show a

clear pattern, consistent acro ss destinations: deteriorating political conditions have a

significantly stronger impact on people with no education or primary education than on

their more educated counterparts. On the contrary, economic conditions do not interact

27

with education. In other words, the im pact of deteriorating or im proving econom ic

conditions does not vary significantly by level of education.

-- Table 4 about here --

As shown in the f irst model (all destin ations), the rate r atio f or the im pact of

political troubles on the less educated is equal to 1.83: an increase of one standard

deviation in the index of political troubles is associated with an 83% increase in the rate

of first m igration. In contrast, political tr oubles are not associated with a significant

increase in migration among people with higher educ ation (rate ratio equal to 1.17, not

significant). A similar (but less pronounced) result is found fo r migration to Africa. On

the other hand, the interaction betw een educat ion and political troubl es is very strong

and highly significant for m igration to W estern countries. In periods of political

troubles, rates of first migra tion increase m ore than thre efold among the less educated,

while they change m ore slightly among the more educated (rate ra tio equal to 1.30, not

significant).

Another way of interpreting this interaction is to look at the impact of education in

period of political crises vs. periods of relative political stability. As shown on Figure 4,

rates of m igration are much higher in periods of intense political crises (corresponding

to the average plus two standard deviations of the indicators) than in periods of political

stability. This figure also clea rly shows that the im pact of crises is stronger am ong the

less educated than among the more educated (measured by the ratio of the black bar to

the grey bar). In other words, m igration is much less selective with respect to education

in periods of political troubles th an in periods of political stability. However, as is also

28

clear from this figure, the more educated remain much more likely to migrate than their

less educated counterparts, even in periods of political troubles.

-- Figure 4 about here --

These m odels show th at, as th e political situation dete riorates, the propensity of

migration increases essentially among the less educated, and very little among the more

educated, resulting in less sele ctive migration. Contrary to what was observed in Peru

(Massey and Capoferro, 2006), selectivity by education decreases as a result of political

troubles, but does not depend on economic conditions. Again, the distinction between

economic and political conditions may not be clear-cut, as these variables are correlated,

but these results at least indicate a strong impact of political con ditions on the

selectivity of migration.

One possible explanation is that the less e ducated as well as the m ore educated are

affected by political crises, and tend to f lee violence and insecurity, while in periods of

stability, the m ore educated are m uch more likely to seek to m igrate. In other words,

this supports the idea that, during crises, people tend to flee deteriorating conditions

(Massey and Capoferro, 2006; Su mata, 2002a), and that m igrations depend to a lesser

extent on h uman capita l in such p eriods than in periods of stability. Another possible

explanation (m aybe com plementary) is th at in tim es of politica l crises – legal

possibilities of m igration are less selective with respect to education. In tim es of

stability, the less educated usually have less legal poss ibilities of migrating than their

educated co unterparts, who can more eas ily get visas for highe r education and for

family reunification. In contrast, asylum applications – which is one of the major way of

legal entry in Western countries among the Congolese since the 1990s with the political

29

troubles in DR Congo – is open to the less educ ated as well as to the m ore educated. A

third possible reason for the lowe r selectivity of m igration according to education may

be due to the fact that bi lateral cooperation seriously slowed down with political

troubles. As a result, a decrease in the number of scholarships granted to Congolese

students may have had an offsetting effect among the better educated on the increasing

desire to migrate in period of political troubles.

8. Discussion and Conclusion

Using recen t retrospective data and even t histo ry m odels, we have shown that

international migration from DR Congo since the mid-1970s has been clearly influenced

by political troubles and, to a lesser extent, by econom ic crises. Periods of political

instability and wars have contributed to significantly higher ri sks of m igration,

especially to Europe and North Am erica, but also to Africa. The graphical analys es and

the robust results of event hist ory models show clearly that international migration and

political crises are closely related in DR Congo. This expected impact of political crises

is in line with resu lts in other parts of the world (Moore and Shellm an, 2004; Stanley,

1987), and suggests th at im proving political conditions in departure countries may

contribute to stabilizing migration f lows. The recent s lowdown of migration from DR

Congo (since 2002) illustrates this influenc e of the improvem ent of the political

conditions in DR Congo on migration.

Economic troubles have also increased m igration, but their effect is less clear-cut

than the impact of political troubles. Results suggest that, when controlling for political

conditions in the models, deteriorating economic conditions only increase m igrations to

Africa. In contrast, migra tions to Europe seem largely unaffected by econom ic

conditions, and have varied e ssentially with the political situation in DR Congo. One

30

possible explanation is that, in pe riods of crises, the availability of resources to migrate

to distant places m ay not be available. A lthough desires to m ove m ay increase, the

financial resources m ay be lacking. This result should be qualified however. First,

economic troubles and political tro ubles are not independent from each other. Mo dels

including only econom ic indicators show that econom ic troubles increase m igration

when all d estinations are cons idered toge ther (the co efficient is a lso positive f or

migration to W estern countries, but not signi ficant). In other words, m igrations do

increase in periods of economic crises, but the association between economic conditions

and migration is very much reduced when political conditions are taken into account.

Secondly, the indicator of econom ic troubles that was used (GDP growth) m ay also

partly explain that the impact of econom ic conditions is no t sign ificant. More refined

indicators might lead to str onger results. However, at this stage, it seems reasonable to

say that political conditions clearly have a strong im pact on m igrations, while the

effects of economic conditions are less conclusive.

Models have also show n that educated and uneducated people respond in different

ways to crises (Figure 4). A consistent result across all the models is that political crises

have a significantly stronger im pact on migration among the less educated than am ong

the more educated, and the im pact is m uch stronger on m igration to Europe and North

America th an on m igration to Africa. This means that, in tim es of political crises,

migration to W estern countries is m uch less selective by ed ucation than in periods of

relative stability. This might be interpreted, as in Massey and Capoferro’s study in Peru

(2006), as the result of different m igration strategies prevailing in periods political

troubles compared to more stable periods. This m ay also reflect changes in legal

possibilities of m igration to W estern count ries in tim es of cr ises – with m ore

31

possibilities of legal entry f or the less educa ted than in periods of stability- and a

decrease in num ber of scholarships for Congolese students in W estern countries,

offsetting a possibly greater desire to move to Western countries.

Further work is needed on several issues . The m echanisms by which political and

economic crises influence m igrations have not been dealt with in de tail in this paper.

The analysis of in-depth interv iews c onducted with Congolese m igrants will prov ide

further insight into the m otivations of m igrations in tim es of crisis. Retrospective

quantitative data collected in Kinshasa among non m igrants may also help understand

the way crises influence m igration. Thes e data include infor mation on m igration

attempts, which m ake it possible to estim ate if increasing migrations in tim es of crises

reflect increasing desires to m igrate (highe r migration attempts), or a higher “success

rate” of migration attempts (i.e. higher probability that an attempt is transformed into an

effective m igration). Finally, retrospectiv e quantitativ e data recen tly collected am ong

Congolese migrants in Belgium and the UK as part of the MAFE project will also shed

light on the dynamics of Congolese migration in times of crises.

References Addleton J. 1991. The impact of the Gulf war o n migration and remittances in Asia and

the Middle East, International Migration, 29(4): 509-26.

Adepoju A. 1994. Prelim inary Analysis of Em igration Dyna mics in Sub-Saharan

Africa, International Migration, 32(2): 197-216.

Akokpari, John, 1999. The political econom y of m igration in Sub-Saharan Africa,

African Sociological Review, 3(1): 15.

Allison P. 1995. Survival Analysis using the SAS System, SAS Institute, Cary, p.292.

32

Bamba, D. L. 2003. La démystification du Parti-Etat au Zaïre. Contribution à uns

dynamique de renouveau démocratique au Congo, PhD Di ssertation, Université

catholique de Louvain, Louvain-la-Neuve, p.277.

Bazenguissa-Ganga, R. 2005. Democratic Republic of Congo (Congo-DRC) and

Republic of Congo (Congo) Country Study. A part of the report on Informal

Remittance Systems in Africa, Caribbean and Pacific (ACP) countries, ESRC

Centre on Migration, Policy and Society (COMPAS), Oxford, p.22.

Blossfeld H.-P., Rohwer G. and G olsch K. 2007. Event History Analysis with Stata,

Lawrence Erlbaum, Philadelphia, p.300.

Conti C. and M. Mam olo 2007. Yougoslavie et As ie centrale : terres de conflits et de

migrations, in AIDELF, Les migrations internationales. Observation, analyse et

perspectives, Presses Universitaires de France, Paris: 499-509.

Crush J. 2008. South Africa: Poli cy in the Face of Xenophobia, Migration Policy

Institute Country Profiles. (http://www.migrationinformation.org).

Dunn, K.C. 2002. A sur vival guide to Kinshasa . Lessons of the father, passed down to

the son, in Clark J. F . (ed), The African Stakes of the Congo War, Fountain

Publishers, Kampala: 53-74.

Henry S., Schoumaker B. and C. Beauchemin. 2004. The Impact of Rainfall on the First

Out Migration: A Multi-level Event- History Analysis in Burkina Faso, Population

and Environment, 25(5): 423-460.

Hesselbein G. 2007. The rise and decline of the Congolese State. An analytical narrative

of state-m aking, Crisis States Research Centre, London School of Economics,

Working Paper, n°21, London School of Economics, London, p. 78.

33

Guo, G. 1993. Event-History Analys is for Left-Truncated Data, Sociological

Methodology, 23: 217-243

IMF, 2002. IMF Approves US$750 m illion PRGF Arrangem ent for t he De mocratic

Republic of the Congo, Press Release No. 02/27 June 13, 2002, In ternational

Monetary Fund, Washington D.C.

Jokisch B. and J. Pribilsky. 2002. The Panic to Leave: Econom ic Crisis and the ‘New

Emigration’ from Ecuador, International Migration, 40(4): 75-101.

Jones R. 1989. Causes of Salvadoran Migration to the United States, Geographical

Review, 79(2): 183-194.

Kadima D., 1999, Motivations à l’ém igration et activités économ iques des imm igrés

congolais (RDC) en Afrique du Sud, in Bouillon A. (ed.), Immigration africaine en

Afrique du Sud. Les migrants francophones des années 1990, IFAS/Kartha la,

Johannesburg/Paris: 105-124

Kagne B. and Martiniello M. 2001. L’imm igration sub-saharienne en Belgique,

Courrier hebdomadaire du CRISP, n°1721, p.50.

Marsh L. and Cor mier D. 2001. Spline regression models, Quantitative Applications in

the Social Sciences, Sage, Thousand Oaks, p.80.

Martin, P. 2009. “Recession an d m igration: A new era for lab or m igration?”,

International Migration Review, 43(3): 671-691.

Massey D. and C. Capoferro. 2006. Sálvese Quien Pueda:Structural A djustment and

Emigration f rom Lim a, The Annals of the American Academy of Political and

Social Science, 606(1): 116-127.

McCalpin J. 2002. Historicity of a c risis. The origins of the Congo W ar, in Clark J. F.

(ed), The African Stakes of the Congo War, Fountain Publishers, Kampala: 33-50.

34

Merli G. 1997. Estim ation of international m igration for Vietnam , 1979-1989, Center

for Studies in Demography and Ecology, University of Washington, Seattle, p.59.

Migration Policy Institute. 2007. Migration Po licy Institute Data Hub, Migration P olicy

Institute, Washington D.C., online database (http://www.migrationpolicy.org).

Moore W. and S. Shellm an. 2004. Fear of Persecution. Forced Migration, 1952-1995,

Journal of Conflict Resolution, 40(5): 723-745.

Morrison A. 1993. Violence or Econom ics: W hat drives internal m igration in

Guatemala?, Economic Development and Cultural Change, 41(4): 817-831.

Morrison A. and May R. 1994. Es cape from t error: violence and m igration in post

revolutionary Guatemala, Latin American Research Review, 29(2): 111-132.

Mutamba Lukusa G., 1999. Congo/Zaïre, la faillite d’un pays. Déséquilibre macro-

économique et ajustements (1988-1999), Cahiers africains, n°37-38,

CEDAF/L’Harmattan, Tervuren/Paris, p.190.

Ndaywel è Nziem , I. 1998. Histoire générale du Congo. De l’héritage ancien à la

République démocratique, De Boeck/Larcier, Bruxelles/Paris, p.955.

Ngoie Tshibam be G. a nd Vwakyanakazi M. 2008, Country Paper: the Dem ocratic

Republic of Congo. Profil m igratoire par pays: cas de la RDC, Perspectives

africaine sur la mobilité humaine, Université de Lubumbashi, Lubumbashi, p. 48.

Nzisabira, J. 1997. Gaspillage technologique, récession économique, expansion de la

pauvreté et dégradation de l’environnement au Zaïre, Institute for New

Technologies, United Nations University, Maastricht, p.175.

PITF. 2009. Political Instability Task Force. Internal Wars and Failures of Governance

1955-2006 (Problem Set Annual Data), online database, consulted on May 2, 2009

(http://globalpolicy.gmu.edu/pitf/pitfpset.htm).

35

Peemans, J. -P. 1997. Le Congo-Zaïre au gré du 20ème siècle. Etat, économie, société

1880 1990, collection Z aïre-Histoire et Sociét é, L’Harm attan, Paris et Montréal,

p.279.

Putzel J., Lindem ann S. a nd C. Schouten, 2008. Drivers of change in the De mocratic

republic of Congo. The rise and declin e of the state and challenges for

reconstruction. A literature review, Crisis States Research Centre, London School

of Economics, Working Paper, n°26, London School of Economics, London, p.43.

RDC and PNUD 2000. Rapport National sur le Développement Humain 2000.

Gouvernance pour le développement humain en RDC, UNDP, Kinshasa, p.227.

Russell S. 1992. International m igration a nd political turm oil in the Middle East,

Population and Development Review, 18(4): 719-727.

Smith D. 1992. Conflicts and W ars, in George, S. (ed), The Debt Boomerang: How

Third World Debt Harms us All., Pluto, London.

Stanley W . 1987. Econom ic Migrants or Re fugees from Violence? A Ti me-Series

Analysis of Salvadoran Migration to the United States, Latin American Research

Review, 22(1), pp. 132-154.

Schoonvaere Q. 2009. Etude de la migration congolaise et de son impact sur la

présence congolaise en Belgique. Analyse des principales données

démographiques, Master dissertation, Université catholique de Louvain, Louvain-

la-Neuve, p.127.

Schoumaker B., Mangalu J. and S. Vause. 2009. Crises et migrations internationales en

RD Congo, Research report, F SP program « Migrations internationales,

recompositions territoriales et dével oppement », Université catholique de

Louvain/Université de Kinshasa, Louvain-la-Neuve/Kinshasa, p. 191.

36

Steinberg, J. 2005. A mixed reception. Mozambican and Congolese Refugees in South

Africa, ISS Monograph Series, n°11 7, Institu te for Security Studies, Cape Town,

p.45.

Stengers, J. 1989. Congo. Mythes et réalités. 100 ans d’histoire, Duculot, Par is-

Louvain-la Neuve, p.283.

Sumata, C. 2002a, Migradollars and poverty alleviation issues in Congo (DRC), Review

of African Political Economy, 29(93): 619-628.

Sumata, C. 2002b, Risk avers ion, internationa l m igration and rem ittances: Congo lese

refugees and asylum seekers in We stern countries, UNU/ WIDER Conference

“Poverty, International Migration and Asylum”, September 2002, Helsinki, p.27.

Sumata, C., T. Trefon a nd S. Cogels, 2004. Images et usages de l’argent de la diaspora

congolaise: les transferts comme vecteur d’entretien du quotidien à Kinshasa, in T.

trefon (ed.), Ordre et d ésordre à Kinshasa : réponses pop ulaires à la faillite de

l’Etat. Tervuren/Paris, Institut africain/L’Harmattan: 135-154.

UNDP. 2009. Statistics of the Human Development Report, online database, consulted

on March 23, 2009 (http://hdr.undp.org/en/statistics/).

UNHCR, 2010a. UNHCR Statistical Online Population Database, Online database,

consulted on January 12, 2010 (http://apps.who.int/globalatlas/default.asp).

UNHCR, 2010b. UNHCR Statistical Online Population Database: sources, methods

and data consideration, online paper, consulted on January 12, 2010

www.unhcr.org/45c06c662.html# refugees

United Nations. 2009. World Population Prospects. The 2008 revision. United Nations,

New York.

37

World Bank. 2009. World Development Indicators, online database, The W orld Bank,

Washington DC.

38

Table 1 : Broad periods in Congo’s political and economic history Period Political situation Economic situation 1960-1965 Political instability after independence Stagnation 1965-1974 Mobutu seizes power . Relative

political stability Growth of the economy

1975-1982 Shaba wars in the late 1970 Deterioration of the economy

1983-1989 Relative stability Economic reform s and slow growth

1990-1996 End of cold war , de mocratization process and start of serious politi cal instability. Riots in the ea rly 1990s, start of the first Congo war in 1996.

Economic deter ioration, negative gr owth r ates, decrease in intern ational development aid.

1997-2002 Regi me change (Mobutu replaced by L-D Kabila) , fir st and second Congo wars. Assassinatio n of L.-D. K abila in 2001 replaced by his son J. Kabila.

Negative growth rates

2003-2007 End of t he war , election in 20 06 of Joseph Kabila.

Improvement of the economic situation, resumption of international aid

39

Table 2. Event-history models of first international migration, Kinshasa (DR Congo), 1975-2007 (results expressed as rate ratios).

Variables Categories Models

1 2 3

Education No / primary

Secondary

Higher education

1.00

2.16***

2.33***

1.00

2.24***

2.50***

1.00

2.23***

2.48***

Gender M ales

Females

1.00

0.77**

1.00

0.78**

1.00

0.77**

Marital status Ever married

Never married

1.00

2.10***

1.00

2.06***

1.00

2.10***

Year (exponential trend) 1.02 (ns) 1.05*** 1.02 (ns)

GDP growth (a) 0.89*** 0.97 (ns)

Political troubles (b) 1.27 *** 1.25***

Sample size 4485 4485 4485

Notes:

a) average GDP growth rate over the two previous years (standardized variable)

b) average index of political troubles over the two previous years (standardized variable)

Significance: *: p<0.10; **: p<0.05; ***: p<0.01 ; (ns) : not significant.

40

Table 3. Event-history models of first international migration by destination of migration, Kinshasa (DR Congo), 1975-2007 (results expressed as rate ratios).

Variables Categories Destination Europe & N.A. Africa Education No / primary

Secondary Higher education

1.00 4.98*** 11.62***

1.00 1.69** 1.28 (ns)

Gender M ales Females

1.00 1.58***

1.00 0.57***

Marital status Ever married Never married

1.00 2.41***

1.00 1.77***

Year (linear trend) 0.97** 1.07*** GDP growth (a) 1.03 (ns) 0.89*** Political troubles (b) 1.41*** 1.13* Sample size 4485 4485 Notes: a) average GDP growth rate over the two previous years (standardized variable) b) average index of political troubles over the two previous years (standardized variable) Significance: *: p<0.10; **: p<0.05; ***: p<0.01 ; (ns) : not significant.

41

Table 4. Event-history models of first international migration by destination, including interactions between education and economic and political troubles, Kinshasa (DR

Congo), 1975-2007 (rate ratios). Destination

Variables Categories All destinations Europe & N.A. Africa

Education No / primary Secondary Higher education

1.00 2.82*** 3.21***

1.00 9.36*** 23.47***

1.00 2.07** 1.59**

Gender M ales Females

1.00 0.78*

1.00 1.57***

1.00 0.57***

Marital status Ever married Never married

1.00 2.05***

1.00 2.45***

1.00 1.79***

Year (linear trend) 1.02* 0.97*** 1.07***

Interactions

GDP gr owth ( a) * Education

No /primary Secondary Higher education

1.09 (ns) 0.93 (ns) 1.03 (ns)

1.05 (ns) 1.05 (ns) 1.00 (ns)

0.99 (ns) 0.85 (ns) 1.00 (ns)

Political troubles (b) * Education

No /primary Secondary Higher education

1.83*** 1.24 (ns) 1.17 (ns)

3.51*** 1.47 (ns) 1.30 (ns)

1.46 ** 1.12 (ns) 1.00 (ns)

Sample size 4485 4485 4485 Notes: a) average GDP growth over the two previous years (standardized variable) b) average index of political troubles over the two previous years (standardized variable) Significance: *: p<0.10; **: p<0.05; ***: p<0.01 ; (ns): not significant.

42

Figure 1. Indices of economic conditions (GDP growth) and political troubles,

DR Congo, 1975-2007.

-15

-10

-5

0

5

10

1970 1975 1980 1985 1990 1995 2000 2005 2010

GD

P gr

owth

(%)

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

Inde

x of

pol

itica

l tro

uble

sGDP grow th

Political troubles

43

Figure 2. ‘Lifetime risk’ of international migration, Kinshasa (DR Congo), 1975-2007.

0%

10%

20%

30%

40%

50%

60%

1970 1975 1980 1985 1990 1995 2000 2005 2010

Year

Life

time

mig

ratio

n ris

k (1

5-50

)

44

Figure 3. Comparison of migration rates and (a) index of political troubles and (b) GDP growth rate, DR Congo, 1975-2007.

0.0%

10.0%

20.0%

30.0%

40.0%

50.0%

60.0%

1972 1977 1982 1987 1992 1997 2002 2007

Year

Life

time

risk

of m

igra

tion

(15-

50)

-3

-2

-1

0

1

2

3

4

5

Inde

x of

pol

itica

l tro

uble

s

Political troubles

Migration

(a) Index of political troubles

0.0%

10.0%

20.0%

30.0%

40.0%

50.0%

60.0%

1972 1977 1982 1987 1992 1997 2002 2007

Year

Life

time

risk

of m

igra

tion

(15-

50)

-15

-10

-5

0

5

10

GD

P G

row

th ra

te (%

)GDP Growth rate (inverse scale)

Migration

(b) GDP growth rate

Note: both indices are computed as the average of the original indices over the two preceding years (e.g. the index represented for the year 1995 is the average for the years 1993 and 1994).

45

Figure 4 . Impact of economic and political troubles on migration to Western countries, by level of education.

0

5

10

15

20

25

30

35

40

45

No education /primary

Secondary education Higher education

Rat

e ra

tios

Average level

Intense crises (+ 2 std dev)

46

1 We use the current name of t he country (Democratic Republic o f Congo, abbreviated as DR Congo)

throughout this article. Between 1971 and 1997, the country was known as Zaïre.

2 According to the UNHCR, more than 300 000 Congolese refugees were living in neighboring countries

in the early 2000s, and around 220 000 in 2008 (UNHCR, 2010a).

3 Various data sources are used to estimate populations of refugees, and, as stated by the UNHCR (2010b,

no page nu mber), “i n m ass r efugee situ ations and when populations are h ighly mobile, m aintaining a

refugee register is a serious challenge”.

4 And when they are available, they are not necessarily comparable across countries.

5 For instance, some migrants can take several months or years to reach their ‘final destination’.

6 Research h as also b een done o n th e im pact o f cr ises in d estination co untries on m igration ( see for

example Martin, 2009), but is not directly relevant to this paper.

7 Research of relevance t o our work has als o bee n done on in ternal m igration. For i nstance, Morris on

(1993), and Morrison and May (1994) compare the impact of economic factors and vi olence on interna l

migration, and con clude th at “ev en when political v iolence is exp licitly introduced, th e co efficients on

purely eco nomic vari ables c ontinue t o be st rongly si gnificant, a nd t he el asticity of migration response

with resp ect to econ omic variables is sig nificantly la rger th an it is with resp ect to v iolence

variables (Morrison 1993, p. 828).

8 Although Kinshasa was not the centre-stage of the wars in the late 19 90s, it was nevertheless prone to

violence. One of i ts manifestations was “a brief witch-hunt against all eastern Congolese and perceived

foreigners, often judged e xclusively by facial appea rances” ( Dunn, 2 002, p.67). T his was s parked by

Laurent-Désiré Kabila’s portraying of the second rebellion in 1998 as a foreign invasion by Rwanda and

Uganda (Dunn, 2002).

9 The s urvey used in this research was funded by the F rench Government through the FSP progr am

« Migrations internationales, recompositions territoriales et développement ». The design of the survey

and of the questionnaires was done by the University of Louvain and University of Kinshasa, in close

collaboration with INED in France and IPDSR in Senegal. The MAFE program (mafe.site.ined.fr) is now

funded by the European Union, and i nvolves three African countries (DR Co ngo, Senegal, and Ghana)

and six European countries (France, Belgium, Italy, Spain, The Netherlands and United Kingdom).

10 The sam pling f rame of t he 2007 Demographic an d Hea lth S urvey was u sed t o sel ect ra ndomly 29

primary sam pling units ( neighbourhoods) in Ki nshasa. Fo ur st reets (secondary sam pling units) w ere

selected randomly in each neighbourhood, and 8 househo lds (tertiary sa mpling units) were selected in

each street. Overall, 943 households were successfully interviewed.

11 The data sets are available at http://globalpolicy.gmu.edu/pitf/pitfpset.htm.

47

12 http://globalpolicy.gmu.edu/pitf/pitfpset.htm, consulted on May 11, 2009.

13 Complex sample desi gn (stratification, m uti-stage sampling, wei ghting) is taken int o account in th e

analyses. Sta ndard e rrors of coeffi cients a re correcte d for t he cluste ring of observations i n the same

neighborhoods.

14 Although age 18 could be more appropriate for migrations to Western countries, migrations to other

African countries (e.g. Angola) can be done alone at younger ages. Results are not sensitive to this choice.

15 The p- value f or a variable to enter the model was set to 0. 09 (Wald t est), and the p- value fo r the

variable to remain in the model was set to 0.10.

16 Rates of first mig ration b y age were transformed into pro babilities o f first m igration by ag e. Th ese

probabilities are then transformed into probabilities of surviving to migration between age 15 and age 50.

The ‘lifetime risk’ of migration is computed as one minus the probability of surviving to first migration at

age 50 (rates of first migration are negligible beyond age 50). A lifetime risk of, say, 30% indicates that if

the rates of migration by age observed in a specific year (e.g. 2007) were constant, 30% of a generation of

people would migration at least once between age 15 and age 50.